Opencv kmeans predict equivalent

Question

I'm moving my project from Python's libraries to opencv, and I have one big problem.

In Python skylearn, I have kmeans object, which has two useful methods: fit_predict and fit, which both are nescessery for my project.

The problem is I can't find any equivalent in opencv. There is only function cv::kmeans, which is not enough for me.

When I'm using Python skylearn the pseudo code of my program looks like:

kmeans.fit_predict(train_data) // it gaves me labels, and compute centers

kmeans.predict(test_sample) // it gaves me labels for test sample computed from centers which were computed above

Is there any equivalent or way to do the same in opencv?

Answer 1

The OpenCV kmeans function is equivalent to fit(). It returns the cluster centers, and you can use this to implement your own predict()function. For a new sample, calculate the distance to each of the cluster centers. The cluster with the lowest distance is the predicted class.